GPT: AI converts brain activity into drawings

GPT: AI converts brain activity into drawings

The artificial intelligence continues to expand into more fields, and its uses are only perplexing everyone. Recently, researchers at Osaka University in Japan developed a mechanism and a method that allowed them to read people’s brain activity using artificial intelligence. AI generative Stable Diffusion to convert this information into drawings.

As the study of the Graduate School of Frontier Biosciences published last December 2023, researchers showed various images to a group of volunteers to stimulate their brain activity that could be captured by MRI scans. The data were translated using semantic decoders and were used to create a final image with the Stable Diffusion.

The scientists noted that this was the first time they had used a tool from generative artificial intelligencea name used to define those capable of creating their own original content, in an experiment to analyze people’s minds and create visual interpretations of such activity.

How does this AI work?

The team of researchers attached to their study a comparison between the images shown to those who participated and those that were finally generated with Stable Diffusion. While they are not completely faithful to the originals, they retain most of the salient features of the samples such as shapes, colors, and perspective of the shot. Of course, the final images also varied among the volunteers because everyone has a different perception of reality.

It should be made clear that the researchers did not only use the IA Stable Diffusion since other processes were also involved to transform brain activity into information. Nevertheless, the platform was able to simplify the creation of graphical representations with good resolution and semantically faithful.

That said, the scientists acknowledged that they took previous studies from other entities such as the Radboud University in the Netherlands to generate the very high resolution photographs. “It was all a process of training and fine-tuning generative models.”the study states.

The Stable Diffusion function

Stable Diffusion Is a generative artificial intelligence created by the company. Stability AI. This platform is able to design images from a text description and, unlike other similar tools such as DALL-E from OpenAIis open source. However, the detail by which it is more useful than its congeners for this type of studies is that it has a machine learning logic. That is, it generates better results the more it is used.

Unfortunately, not everything is rosy for this platform since, last January, it was sued by a trio of artists who accused it of having violated the copyrights of “millions of artists”. Getty Images also sued her in February claiming he used millions of her photographs without permission.

We recommend METADATARPP’s technology podcast. News, analysis, reviews, recommendations and everything you need to know about the technological world.


Daniel Chapman